Related Experiment Video
Updated: Aug 30, 2025
06:52
Discovery of Driver Genes in Colorectal HT29-derived Cancer Stem-Like Tumorspheres
Published on: July 22, 2020
6.6K
Identifying cancer tissue-of-origin by a novel machine learning method based on expression quantitative trait loci
Yongchang Miao1,2,3, Xueliang Zhang4, Sijie Chen5
1Gastroenterology Center, The Second People's Hospital of Lianyungang, Lianyungang, China.
Frontiers in Oncology
|August 26, 2022
Summary
Identifying the tissue-of-origin (TOO) for cancer of unknown primary (CUP) is crucial for effective treatment. This study developed a novel computational method integrating expression quantitative trait loci (eQTL) to accurately predict CUP
Area of Science:
- Genomics
- Computational Biology
- Oncology
Background:
- Cancer of unknown primary (CUP) presents a significant clinical challenge due to unidentifiable primary lesions, leading to poor patient survival.
- Accurate identification of the tissue-of-origin (TOO) is critical for targeted cancer therapy and improved outcomes in CUP patients.
- Existing computational methods for TOO inference often rely on single omics data, potentially missing complex biological interactions.
Purpose of the Study:
- To develop and validate a novel computational method for predicting the tissue-of-origin (TOO) in cancer of unknown primary (CUP) by integrating expression quantitative trait loci (eQTL) data.
- To assess the performance of the proposed eQTL-integrated model against models using single omics data.
- To evaluate the clinical applicability of the developed model for routine CUP diagnosis.
Main Methods:
- Developed a novel computational classification model utilizing XGBoost, incorporating expression quantitative trait loci (eQTL) data.
- Trained and validated the model using The Cancer Genome Atlas (TCGA) dataset, comprising over 7,000 samples across 20 solid tumor types.
- Performed 10-fold cross-validation and tested the model on an independent Gene Expression Omnibus (GEO) dataset.
Main Results:
- The eQTL-integrated XGBoost model achieved a prediction accuracy exceeding 0.96 in 10-fold cross-validation on TCGA data.
- The model demonstrated superior performance compared to methods not incorporating eQTL data.
- On an independent GEO dataset, the model achieved an F1-score ranging from 0.7 to 1.0, varying by cancer type.
Conclusions:
- Expression quantitative trait loci (eQTL) provide crucial information for accurate tissue-of-origin inference in cancer of unknown primary (CUP).
- The developed computational model demonstrates high accuracy and potential for clinical application in diagnosing CUP.
- Integration of multi-omics data, specifically eQTL, enhances the predictive power of computational models for complex diseases like CUP.
